Mobile-Captured Image Document Recognition for Vietnamese Receipts (MC-OCR) Forum

Go back to competition Back to thread list Post in this thread

> is pretrained model or additional data allowed

is pretrained model or additional data allowed

Posted by: haituuuuu @ Dec. 22, 2020, 11:34 a.m.

Hi haituuuuu,

Thanks for asking.
To be fair between all participants, external data is NOT allowed.
We will open a thread to register for the use of public pre-trained models.

Regards,

Posted by: mc-ocr-organizers @ Dec. 22, 2020, 5:59 p.m.

can i self label public and warmup datasets

Posted by: haituuuuu @ Dec. 23, 2020, 1:57 a.m.

All data must be used in As-Is condition.
Manual annotation and extra data are NOT allowed.

Posted by: mc-ocr-organizers @ Dec. 23, 2020, 7:46 a.m.

Please help me to clarify some works that are allowed or not:
1. Modify label from training set
2. Modify polygon from training set
3. Add more polygon + label in training set
4. Do 1, 2,3 for warmup set or test set and use them for training.

Posted by: SDSV_AICR @ Dec. 24, 2020, 2:28 a.m.

Hi SDSV_AICR,

> 1. Modify label from training set (not allowed)
> 2. Modify polygon from training set (not allowed)
> 3. Add more polygon + label in training set (not allowed)
> 4. Do 1, 2,3 for warmup set or test set and use them for training (not allowed)

All of the above items are NOT allowed.
As we emphasised in previous messages, this is an automatic recognition data challenge and manual annotation is NOT allowed.

Posted by: mc-ocr-organizers @ Dec. 24, 2020, 9:27 a.m.

Hi SDSV_AICR,

Yes, rule base methods are acceptable.
As long as you don't have human annotators get involved to modify the training data, you're good to process.

Regards,

Posted by: mc-ocr-organizers @ Dec. 31, 2020, 4:51 p.m.
Post in this thread